[OSM-dev] Anyway to split or simplify an .osm file from the command line?

Frederik Ramm frederik at remote.org
Sun Sep 20 15:30:40 BST 2009


John Smith wrote:
> The majority of Australian postcodes have been converted from shape
> files to osm files, however there is a couple of complex/big areas
> that I'm having trouble dealing with in JOSM as a result, is there a
> simple way to simplify these files from a command line util?

Let me speak bluntly: Have you been in touch with *anyone* who knows 
something about OSM before you attempted this import?

One of the files you are trying to import has a node with more than 
40.000 ways. Even if JOSM would handle it, the API would not accept it. 
Even if teh API would accept it, you would basically sabotage any 
attempt at mapping anything touching the boundary of that post code area 
because they would be forced to download 4 MB of postcode area data even 
if they just wanted a little stretch of road that happened to cross it.

Your ways are not tagged with anything. Did you plan to add the tags 
manually in JOSM before uploading?

I have not checked this but given the ignorance that speaks from the 
above, I very much assume that you have not undertaken any effort to 
re-use nodes and ways; my guess is that where there is a boundary 
between two post code areas, you will import every node twice, once as 
part of post code area 1 and once as part of post code area 2... right? 
This is an awful waste of space and not how we do things in OSM.

Please stop this import, take a deep breath, find out how to do it 
right, and then do it properly.

Could the author(s) of shp2osm also please make sure that users of the 
tool are educated about multipolygons and the maximum sizes of ways.

John, the following thread should have some information about how to 
properly import a mesh of boundaries (which I believe is what you are 
planning to do):


In short, you need to create "multipolygon" relations consisting of ways 
with no more thane 2,000 members each, and where you have shared 
boundaries these should be ONE way with ONE set of nodes, with the way 
referenced by the two post code area multipolygon.

The thread also mentions the newly setup "imports support" working group 
which you can contact for help if you want.

Please, folks, be more careful with imports. I'd rather *not* have post 
code data than botched imports. Well-meaning as they are, they can 
easily do more damage than any vandalism we have seen until now.


More information about the dev mailing list